pnpla1 has a crucial role in skin barrier function by directing acylceramide biosynthesis

We provide evidence that absence of PNPLA1 causes a severe skin permeability barrier defect by perturbing the linoleoyl o-O-esterification of ceramides to yield acylceramides, along with

PNPLA1 has a crucial role in skin barrier function

by directing acylceramide biosynthesis

Tetsuya Hirabayashi 1,2, *, Tatsuki Anjo 1,3, *, Arisa Kaneko 1,4 , Yuuya Senoo 5 , Akitaka Shibata 6 , Hiroyuki Takama 6 , Kohei Yokoyama 1 , Yasumasa Nishito 7 , Tomio Ono 7 , Choji Taya 7 , Kazuaki Muramatsu 3 , Kiyoko Fukami 4 ,

Agustı´ Mun ˜oz-Garcia 8 , Alan R Brash 9 , Kazutaka Ikeda 5 , Makoto Arita 5 , Masashi Akiyama 6

& Makoto Murakami 1,2

Mutations in patatin-like phospholipase domain-containing 1 (PNPLA1) cause autosomal

recessive congenital ichthyosis, but the mechanism involved remains unclear Here we show

that PNPLA1, an enzyme expressed in differentiated keratinocytes, plays a crucial role in the

biosynthesis of o-O-acylceramide, a lipid component essential for skin barrier Global or

keratinocyte-specific Pnpla1-deficient neonates die due to epidermal permeability barrier

defects with severe transepidermal water loss, decreased intercellular lipid lamellae in the

stratum corneum, and aberrant keratinocyte differentiation In Pnpla1 /  epidermis,

unique linoleate-containing lipids including acylceramides, acylglucosylceramides and

(O-acyl)-o-hydroxy fatty acids are almost absent with reciprocal increases in their putative

precursors, indicating that PNPLA1 catalyses the o-O-esterification with linoleic acid to form

acylceramides Moreover, acylceramide supplementation partially rescues the altered

differentiation of Pnpla1 / keratinocytes Our findings provide valuable insight into the skin

barrier formation and ichthyosis development, and may contribute to novel therapeutic

strategies for treatment of epidermal barrier defects.

1Lipid Metabolism Project, Tokyo Metropolitan Institute of Medical Science, Tokyo 156-8506, Japan.2AMED-CREST, Japan Agency for Medical Research and Development, Tokyo 100-0004, Japan.3Division of Life Science and Engineering, School of Science and Engineering, Tokyo Denki University, Saitama

350-0394, Japan.4Laboratory of Genome and Biosignals, Tokyo University of Pharmacy and Life Sciences, Tokyo 192-0392, Japan.5Laboratory for Metabolomics, RIKEN Center for Integrative Medical Sciences, Kanagawa 230-0045, Japan.6Department of Dermatology, Nagoya University Graduate School of Medicine, Aichi 466-8550, Japan.7Center for Basic Technology Research, Tokyo Metropolitan Institute of Medical Science, Tokyo 156-8506, Japan.8Department of Evolution, Ecology, and Organismal Biology, Ohio State University, Ohio 43210, USA.9Vanderbilt Institute of Chemical Biology, Vanderbilt University, Nashville, Tennessee 37232-6304, USA * These authors contributed equally to this work Correspondence and requests for materials should be addressed to M.M (email: murakami-mk@igakuken.or.jp)

T he skin barrier of terrestrial mammals is essential for

prevention of water and electrolyte loss, as well as

protection from the penetration of harmful substances

and pathogenic microbes1,2 Impairment of skin barrier function

can cause or aggravate skin disorders, including dry skin,

ichthyosis, psoriasis, and atopic dermatitis3–7 Although the

epidermis is a highly organized stratified epithelium consisting of

four distinctive layers, the innermost stratum basale (SB), the

stratum spinosum (SS), the stratum granulosum (SG) and the

uppermost stratum corneum (SC), its barrier function is provided

mainly by specialized structures in the SC and tight junctions in

the SG8,9 The unique SC components include cross-linked,

insoluble proteins of corneocytes forming the cornified envelope

(CE) and its associated, external membrane monolayer, called the

cornified lipid envelope (CLE), as well as the intercellular lipid

lamellae, which are mainly composed of ceramides, cholesterol

and free fatty acids (FFAs) and are secreted as lamellar body

lipids by differentiated keratinocytes at the SG/SC interface10–12.

Among the epidermal ceramides with marked molecular

heterogeneity (at least 12 classes in humans)13,14, acylceramide

is essential for physical and functional organization of lipids in

the SC interstices, and thereby the barrier function of the

skin1,3,11,15 Impaired biosynthesis or processing of acylceramide

causes ichthyosis, characterized by dry, scaly and thickened skin.

Acylceramide is an unusual ceramide species whose N-acyl chain

is composed of o-hydroxylated ultra-long chain FAs (ULCFAs)

esterified at the o-position with linoleic acid (LA; C18:2) It has

been suggested that the ULCFA portion of acylceramide spans a

bilayer while the LA tail inserts into a closely apposed section of

bilayer, thus serving as a molecular rivet to link two membranes

together in the lipid lamellae16 In addition, acylceramides

containing o-O-esterified fatty acids other than LA cannot be

converted to covalently protein-bound o-hydroxyceramide

(Cer OS), which forms CLE and functions as a template on the

surfaces of corneocytes for the organization of lipid layers in

the SC interstices Indeed, in essential FA deficiency, LA in

acylceramide is replaced by oleic acid, which fails to support skin

barrier function properly17 Until now, several important steps

for acyceramide biosynthesis and processing in the epidermis

have been identified from studies of autosomal recessive

congenital ichthyosis (ARCI) in humans and corresponding

mouse disease models with genetic knockouts: the synthesis of

ULCFAs by the FA elongase ELOVL4, o-hydroxylation

of ULCFAs by the FA o-hydroxylase CYP4F22 (or CYP4F39 in

mice), and formation of ceramides with ULCFAs by the ceramide

synthase CERS3 (refs 18–20) However, the mechanism

underlying the formation of acylceramide with LA in the

epidermis is still under debate.

The current model for o-O-esterification of ULCFA-ceramides

with LA involves the hydrolysis of triacylglycerol (TG) in lipid

droplets to provide LA via a yet unknown lipase, followed by its

transfer (via acyl-CoA form) to the o-hydroxy group of ULCFA

in ceramides or glucosylceramides (GlcCer) by a putative

o-O-acyltransferase Alternatively, LA can be directly transferred

from TG to o-OH Cer and/or o-OH GlcCer by an

LA-specific transacylase8,21 Recent studies of patients with

neutral lipid storage disease with ichthyosis (NLSDI or

Chanarin-Dorfman syndrome) and in mice with Abhd5

deletion suggest that TG accumulation due to loss-of-function

of ABHD5 (also known as CGI-58) reduces the availability of LA

for acylceramide production22,23 ABHD5 is an essential co-factor

for stimulation of ATGL (adipose triglyceride lipase, also known

as PNPLA2 or iPLA2z), which plays a major role in TG hydrolysis

in most tissues24,25 ATGL is a member of the patatin-like

phospholipase domain-containing protein (PNPLA) or Ca2 þ

-independent phospholipase A2(iPLA2) family, which comprises 9

enzymes in humans acting as lipid hydrolases, acyltransferases or transacylases with diverse substrate specificities including phospholipids and neutral lipids26 Interestingly, ichthyosis features and decreased acylceramide levels in the skin have been observed in patients and mice with defective ABHD5 function, but not in those with ATGL mutations or deletion, leading to the proposal that ABHD5 could activate a different lipase that regulates epidermal TG hydrolysis21,25 Nonetheless, the molecular entity of o-O-acyltransferase or transacylase responsible for the linoleoyl o-O-esterification of ULCFA-ceramides has not yet been identified.

Loss-of-function mutations in PNPLA1, a paralog of ATGL/PNPLA2, have recently been identified in humans or dogs with ARCI (refs 27–29) PNPLA1 fails to hydrolyse TG, however, even in the presence of ABHD5 (ref 29), raising the question of the role of this functionally orphan enzyme How lipid metabolism regulated by PNPLA1 contributes to epidermal homoeostasis is a fundamental issue which remains to be addressed To this end, we herein generated Pnpla1-deficient mice in our ongoing efforts to decipher the biological roles

of PLA2-related enzymes by gene targeting30–33 We provide evidence that absence of PNPLA1 causes a severe skin permeability barrier defect by perturbing the linoleoyl o-O-esterification of ceramides to yield acylceramides, along with abnormal differentiation of keratinocytes, thus demonstrating that PNPLA1 is a long-sought enzyme that plays

a critical role in acylceramide synthesis in the skin.

Results Expression of PNPLA1 in highly differentiated keratinocytes Among adult mouse tissues, Pnpla1 messenger RNA (mRNA) was expressed most abundantly in the skin, followed by the stomach (Supplementary Fig 1a) Immunohistochemistry of newborn mouse skin revealed localization of PNPLA1 protein in the boundary area between the nucleated SG and the denucleated

SC, just above the location of the granular layer marker loricrin,

in the epidermis (Supplementary Fig 1b) In agreement with a previous report29, PNPLA1 was partially colocalized with filaggrin (a SG marker), but not with keratin 1 and 5 (SS and

SB markers, respectively) (Supplementary Fig 1b) In adult mouse skin, the localization of PNPLA1 in the epidermis was essentially the same as that in newborn skin (Supplementary Fig 1c) In a monolayer culture of mouse keratinocytes, Ca2 þ treatment resulted in marked induction of keratinocyte differentiation markers (Krt10 and Lor), as well as Pnpla1 (Supplementary Fig 1d) Likewise, a marked increase of PNPLA1 expression was observed in human keratinocytes after

Ca2 þ-induced differentiation, an event that occurred in parallel with induced expression of the keratinocyte differentiation markers KRT1 and FLG, but not with constitutive expression of the SB marker KRT5 (Supplementary Fig 1e) These results suggest that PNPLA1 has a specific role in highly differentiated keratinocytes in the uppermost layer of the SG, where lipids required for epidermal barrier function are processed and secreted into the intercellular space to form lipid lamellae and CLE.

Impaired epidermal permeability barrier in Pnpla1 /  mice.

To gain insight into the function of PNPLA1 in vivo, we gener-ated mice with targeted disruption of the Pnpla1 gene on a C57BL/6 background (Supplementary Fig 2a,b) The absence of mRNA and protein for PNPLA1 in the skin of Pnpla1 /  mice was confirmed by quantitative PCR (qPCR) (Supplementary Fig 2c) and immunohistochemistry (Fig 1a), respectively Offspring from heterozygote intercrosses were born at the

expected Mendelian ratio (Supplementary Fig 2d) Although

Pnpla1þ /  mice were healthy and indistinguishable from

Pnpla1þ / þ mice, newborn Pnpla1 /  pups had shiny and taut

skin, often with a necrotic tail tip (Fig 1b; Supplementary Fig 2e),

and died within 24 h after birth We hypothesized that the cause

of death in these Pnpla1-deficient mice might be dehydration, and

therefore we assessed their skin permeability barrier function.

Pnpla1 /  pups delivered by Caesarean section at E18.5 had

normal body weight as compared with littermate wild-type (WT)

and heterozygous mice at birth, but rapidly lost as much as 20%

of their weight within 16 h (Fig 1c) In accordance with this steep

weight loss, trans-epidermal water loss (TEWL) was markedly

higher in Pnpla1 /  newborns than in Pnpla1þ / þ and

Pnpla1þ /  newborns (Fig 1d), indicating a severe defect of the inside-out barrier in the null mice In the toluidine blue exclusion assay to assess the outside-in permeability barrier, WT littermates excluded dye, whereas Pnpla1 /  pups showed robust dye penetration into the skin (Fig 1e) These phenotypes, which have been commonly observed in mutant mice with disruption of genes associated with ARCI (ref 1), suggest that PNPLA1 is required for epidermal permeability barrier function.

Histological analysis revealed that control mice had a clear basket weave-like structure segregated by interspaces, indicative

of the presence of lipid lamellae (Fig 1f, left) In contrast, Pnpla1 /  mice exhibited a tightly packed structure in the SC, a reduced number of keratohyaline granules in the uppermost SG,

+/+

PNPLA1/DAPI

P0

+/+

+/+

+/+

+/+

+/+

+/+

+/– –/–

–/–

–/–

–/–

–/–

+/–

–/–

6

–2 h

–1)

4 2 0

1.4 1.2 1 0.8 0.6 0.4 0.2 0

Hours

SC SC

SG SG

+/+

–/–

–/–

a

c

f

g

b

Figure 1 | Impaired skin barrier function in Pnpla1 / mice (a) Immunohistochemical staining for PNPLA1 (red), followed by counterstaining with DAPI (blue), in skin sections from Pnpla1þ / þ and Pnpla1 / newborns Dashed lines indicate the upper border of the epidermis (b) Gross appearance of Pnpla1þ / þand Pnpla1 /  newborns at P0 (c) Monitoring of body weights of Pnpla1þ / þ, Pnpla1þ /  and Pnpla1 / mice (two pups for each) after Caesarean section at E18.5 (d) Skin permeability as assessed by TEWL on the dorsal skin surface of Pnpla1þ / þ(n¼ 19), Pnpla1þ /  (n¼ 21) and Pnpla1 / mice (n¼ 21) (mean±s.e.m., ***Po0.001 in an unpaired, two-tailed Student’s t-test) (e) Toluidine blue exclusion assay using neonatal Pnpla1þ / þand Pnpla1 / mice (f) Histology of dorsal skin sections from newborn Pnpla1þ / þand Pnpla1 / mice stained with hematoxylin and eosin Arrow indicates a basket weave structure in the SC of WT mice Mutant epidermis was notably thicker (double-headed arrows) and had a more tightly packed SC (dashed arrow) than WT epidermis (g) Staining of SC lipids with Nile red Continuous linear lipid structures in the SC of Pnpla1þ / þmice were replaced by a dot-like pattern (yellow arrows) in the SC of Pnpla1 / mice (h,i) Transmission electron microscopy of skins of Pnpla1þ / þand Pnpla1 /  newborn mice Compared with Pnpla1þ / þskin, Pnpla1 / skin displayed numerous lipid aggregates in corneocytes (blue boxes), abnormalites in the secreted contents at the SG–SC interface (red circles) (h), and impaired formation of the CLE (arrows) (i) Scale bars; 20 mm (a,f), 5 mm (g), 0.4 mm (h) and 0.2 mm (i) Representative (a–c,e–i) or complied (d) results from at least three experiments are shown

and epidermal hyperplasia (Fig 1f, right), which is considered to

be an adaptive response to barrier disruption Nile red staining of

the Pnpla1þ / þ epidermis showed wavy lipid multilayers

characteristic of SC intercellular lipid lamellae, whereas

granular-like lipid aggregates were present within increased

number of densely packed lamellar sheets in the Pnpla1 / 

epidermis (Fig 1g), suggesting that keratinocytes are

hyperpro-liferative and defective in the secretion and/or composition of SC

lipids in mutant mice Ultrastructural examination of Pnpla1 / 

mice by transmission electron microscopy confirmed the tightly

stacked layers of corneocytes with a substantially decreased

amount of intercellular lipid lamellae, as evidenced by narrowed

SC interstices, in comparison with control mice (Fig 1h) At the

SG–SC interface of Pnpla1þ / þ epidermis, lipid lamellae were

released into the intercellular spaces from lamellar bodies

(Fig 1h) In contrast, the secretion of lipid granule contents

was hampered and abnormal vesicular structures, which were

thought to represent defective lamellar bodies, were retained

within corneocytes in Pnpla1 /  epidermis (Fig 1h) Moreover,

Pnpla1 /  mice exhibited either loss or abnormalities of the

CLE (Fig 1i) These results indicate that PNPLA1 plays a critical

role in the proper formation of intercellular lipid lamellae and

CLE in the SC, which are important for the permeability barrier

function of the skin.

Aberrant keratinocyte differentiation in Pnpla1 /  mice To

further address the skin abnormalities in Pnpla1 /  mice, we

performed microarray gene profiling using skins of newborn

Pnpla1þ / þ and Pnpla1 /  mice Heat map visualization of

selected genes indicated down-regulation of genes for late

keratinocyte differentiation and CE constituents (for example,

Lor, Flg, Flg2, and the late cornified envelope genes Lce1a, b) in

Pnpla1 /  epidermis (Supplementary Fig 3a) Most of these

are located within the epidermal differentiation complex, a

keratinocyte lineage-specific gene locus on mouse chromosome 3.

Several up-regulated gene clusters within the epidermal

differentiation complex, such as small proline-rich proteins

(Sprr1a, Sprr1b and Sprr2b), late cornified envelope proteins

(Lce3b and Lce3c), and S100 proteins (S100a8 and S100a9), have

been associated with psoriasis34 Other up-regulated genes

included those involved in keratinocyte proliferation linked to

epidermal growth factor (EGF) signalling (for example, Areg,

Epgn, Tgfa, Hbegf and Ereg), adhesive structures (for example,

Cldn4, Cldn7, Dsc1, Dsc2, Ocln, Tjp1 and Tjp2), lipid metabolism

(for example, Fasn, Scd1, Pla2g4d and Pla2g4e) and

skin-associated immune responses (for example, Il1b, Il12a, Il13,

Il20, Il22, Il23a, Tnf, Ifng and Cxcl1) (Supplementary Fig 3a) It is

likely that the enhanced expression of inflammatory cytokines

and chemokines is a secondary effect resulting from impaired

barrier function, since similar changes have also been observed in

several genetically distinct mouse models with barrier defects33–38

and patients with skin diseases such as ichthyosis, atopic

dermatitis and psoriasis39,40 Interestingly, expression levels of

genes associated with synthesis and processing of epidermal

acylceramide were consistently elevated in Pnpla1 /  mice

relative to Pnpla1þ / þ and Pnpla1þ /  mice (Supplementary

Fig 3a,b) These genes included Elovl4, Abhd5, Cers3, Cyp4f39 (a

mouse ortholog of human CYP4F22), Ugcg, Abca12 and Gba,

mutation or deletion of which has been shown to cause ARCI in

humans and neonatal death in mice due to severe skin barrier

defects1,41.

Immunofluorescence staining and qPCR confirmed the

diminished expression of terminal differentiation markers, such

as filaggrin (Flg) and loricrin (Lor), in Pnpla1 /  skin relative to

Pnpla1þ / þ skin, whereas mRNA and protein expression levels of

the basal and early suprabasal keratinocyte differentiation

markers, keratin 5 (Krt5) and 1 (Krt1), were similar between the two genotypes (Fig 2a,b) In contrast, keratin 6 (Krt6a and Krt6b) was expressed in the lower suprabasal layer in Pnpla1-deficient but not in control skin, reflecting the hyperproliferative state of the mutant epidermis Abnormal differentiation of keratinocytes has also been observed in several mouse lines with targeted disruption of genes implicated in epidermal ceramide metabolism22,35 Therefore, the neonatal lethality of Pnpla1 /  mice due to skin barrier defect is likely dependent upon both altered lipid composition and impaired differentiation of keratinocytes.

Moreover, expression of PPARd (Ppard) and its potential target genes such as Fabp5 and Sprr1b (ref 35) was markedly increased in Pnpla1 /  skin relative to WT skin (Fig 2c; Supplementary Fig 3a), indicating that PNPLA1 deficiency leads

to hyperactivation of PPARd Activation of EGF receptors has been shown to control keratinocyte proliferation and differentia-tion with decreased expression of differentiadifferentia-tion-related genes including filaggrin and loricrin42 Indeed, heparin-binding EGF-like growth factor (HB-EGF), a potent autocrine growth factor for keratinocytes and putative target gene of PPARd (ref 43), was robustly upregulated in Pnpla1 /  epidermis (Fig 2c; Supplementary Fig 3a), suggesting that EGF receptor signalling contributes, at least in part, to epidermal hyperplasia and altered keratinocyte differentiation in the mutant mice.

Defective acylceramide biosynthesis in Pnpla1 /  skin To identify the endogenous lipid metabolism regulated by PNPLA1,

we performed thin-layer chromatography (TLC) and quantitative liquid chromatography mass spectrometry (LC–MS/MS) using epidermal lipids extracted from neonatal WT and mutant mice TLC analysis revealed that the bands for acylceramide (esterified omega-hydroxyacyl-sphingosine; EOS), which is a key determi-nant of skin permeability barrier function15, and its derivative acylglucosylceramide (GlcEOS) were markedly reduced or almost undetectable in Pnpla1 /  mice relative to WT and heterozygous mice (Fig 3a) We also noticed that another lipid species, with a TLC motility slightly faster than that

of FA, was nearly absent in mutant mice, and LC–MS/MS analysis with collision-induced fragmentation of this lipid extracted from the TLC plate identified it as (O-acyl)-o-hydroxy FA (OAHFA), particularly (O-linoleoyl)-o-(O-acyl)-o-hydroxy FA (OLHFA) (see below) In contrast, o-hydroxy FA (o-OH FA), o-OH Cer and GlcCer were present in substantially greater amounts in Pnpla1 /  mice than in control Pnpla1þ / þ and Pnpla1þ /  mice (Fig 3a).

To determine the changes in ceramide molecular species in terms of the length and saturation of their N-acyl chains, lipids extracted from Pnpla1þ / þ and Pnpla1 /  epidermis were analysed quantitatively by LC–MS and LC–MS/MS Epidermal ceramide species are grouped into non-hydroxylated ceramides (NS, NDS, NH and NP), a-hydroxylated ceramides (for example,

AS, ADS, AH and AP) and acylceramides (for example, EOS, EOH and EOP), where S, P, DS and H stand for sphingosine, phytosphingosine, dihydrosphingosine and 6-hydroxys-phingosine, respectively44,45 EOS and EOP species with residues of (O-linoleoyl)-o-hydroxy ULCFAs (C28–C38) were almost entirely lost in the epidermis of Pnpla1 / mice (Fig 3b; quantitative data for representative molecular species are depicted

in Supplementary Fig 4a,b) Correspondingly, there was marked accumulation of various molecular species of o-OH Cer, a putative precursor of EOS, in mutant mice relative to WT mice (Fig 3c) In addition, in mutant mice, the amounts of OLHFA species with C28–C36 ULCFAs were markedly decreased, with reciprocal increases in corresponding o-OH ULCFA species (Fig 3d,e), confirming the results of TLC analysis (Fig 3a).

Moreover, the amount of Cer OS covalently bound to the CE was

robustly reduced in mutant mice relative to WT mice

(Supplementary Fig 4c) In contrast to the dramatic reductions

of acylceramides and their downstream products, various

ceramide molecular species (AS, AP, NS, NH and NP) were

modestly increased in Pnpla1 /  mice (Supplementary

Fig 4d–h) Collectively, these data suggest that PNPLA1 is

required for linoleoyl o-O-esterification of the free and/or

ceramide-bound forms of o-OH ULCFA residues Interestingly,

the linoleate residue of several, if not all, EOS and OAHFA

species was partially replaced by the palmitate or oleate residue in

Pnpla1 /  mice (Fig 3b,d), indicating that, in the absence of

PNPLA1, another putative acyltransferase or transacylase with

weak activity and broad substrate specificity may contribute to

the synthesis of EOS and OAHFA with non-linoleate fatty acid

(that is, palmitate or oleate).

Although the epidermal levels of total ceramides were similar

in both genotypes, those of FFAs, cholesterol and TG were

substantially higher in Pnpla1 /  mice than in Pnpla1þ / þ

mice (Fig 3a; Supplementary Fig 5a) Since a proper ratio of

ceramides, FFAs, and cholesterol is crucial for formation of the

SC lipid lamellae, altered proportion of these lipids may lead to their unusual aggregation, as seen in the Pnpla1 /  SC (Fig 1g) Among the FFAs, the levels of very long chain FAs (VLCFAs;

Z C22:0), but not those of long chain FAs (LCFAs), were increased in Pnpla1 /  mice (Supplementary Fig 5b–d) These increases in cholesterol, VLCFAs and several ceramide species resulting from Pnpla1 deficiency accorded with the elevated expression levels of genes related to lipid metabolism such as Hmgcr, Elovl4 and Degs2 (Supplementary Fig 3a,b), suggesting compensatory adaptation of the Pnpla1 /  epidermis to the impaired acylceramide synthesis and barrier formation More-over, our observation that the free LA level was unchanged in Pnpla1 /  mice (Supplementary Fig 5d) argues against the alternative idea that PLPLA1 acts as a TG lipase that supplies LA for o-O-esterification of ULCFA Although the composition of phospholipids was not profoundly affected by Pnpla1 deficiency, some phosphatidylethanolamine (PE) species with polyunsatu-rated fatty acids, including LA, were present in slightly greater amounts in Pnpla1 /  than in Pnpla1þ / þ mice

Filaggrin

+/+

+/+

Flg

Lor

Krt1

Krt5

Krt6a

Hbegf Fabp5

Ppard

–1)

Krt6b

4

*

***

***

**

***

***

***

2

0 20

10

10 0

5

0 10

10 5

5 0

0

30 20

20 60

60

40 20 40

20

0 0

10

10

0

+/+

–/–

0

+/–

–/–

–/–

Loricrin

Keratin 1

Keratin 5

Keratin 6

a

c

b

Figure 2 | Aberrant terminal differentiation of Pnpla1 / epidermis (a) Immunohistochemical staining of keratinocyte differentiation markers (red), followed by conterstaining with DAPI (blue), in skin sections from Pnpla1þ / þand Pnpla1 / newborn mice Scale bars, 20 mm (b) qPCR analysis of keratinocyte differentiation markers in newborn Pnpla1þ / þ, Pnpla1þ / and Pnpla1 / epidermis (n¼ 5 animals per group) (c) qPCR analysis of PPARd (Ppard) and its potential target genes in newborn Pnpla1þ / þ and Pnpla1 /  epidermis (n¼ 7 animals) In b,c, values are mean±s.e.m.; *Po0.05,

**Po0.01, and ***Po0.001 versus Pnplaþ / þmice Representative results from two or three independent experiments are shown

OAHFA

ω-OA FA

ω-OH Cer

GlcEOS

GlcCer

PE

PC

SM

300,000

OAHFA

Cer(EOS)

a

d

e

c

b

+/+

+/+

+/+

+/+

–/–

–/–

–/–

–/–

–/–

ω-OH Cer

ω-OH ULCFA 250,000

250,000

200,000

200,000

150,000

150,000

1,200,000

100,000

100,000

1,000,000

1,000,000

800,000

600,000

300,000

200,000

100,000

0

32:0/16:0 32:0/16:1 32:0/18:0

28:0 30:0 32:0 32:1 33:0 34:0 34:1 34:2 35:0 35:1 36:0 36:1 36:2 38:1 38:2 40:1 30:0/18:2 30:0/18:3 32:0/16:0 32:0/18:2 32:1/16:0 32:1/18:2 32:1/18:3 33:0/18:2 33:1/16:0 34:0/18:2 34:1/16:0 34:1/18:1 34:1/18:2 34:2/18:2 36:0/18:2 36:1/18:1 36:1/18:2 36:2/18:2 38:1/18:2

26:1 27:0 28:0 28:1 28:3 29:0 29:1 30:0 30:1 30:3 31:0 31:1 32:0 32:1 32:2 33:0 33:1 34:0 34:1 34:2 34:3 35:1 36:0 36:1 36:2 36:3 38:1 38:2 38:3 40:1

33:1/16:0, 34:0/16:0, 34:1/17:0, 34:1/18:0, 32:0/18:2, 33:0/18:2, 34:0/18:2, 34:1/18:2, 34:2/18:2, 35:1/18:1, 36:1/18:2, 32:1/24:0, 33:0/16:0 34:1/15:0 34:1/16:0 35:1/16:0 36:1/16:0 35:1/18:0 36:1/18:0 36:1/18:1 34:1/16:1 34:2/17:0 33:1/18:2 34:1/18:1 34:2/18:1 34:1/18:3 35:0/18:2 35:1/18:2 36:0/18:2 36:2/18:1 37:1/18:2 38:1/18:2 38:2/18:2 40:1/18:2 34:2/18:3 34:1/20:4 32:0/24:0 34:1/22:0 34:2/24:0 36:2/24:0

50,000

50,000

800,000 600,000 400,000 200,000 0

0

**

**

**

**

** ** **

**

**

*

*

*

*

**

***

***

***

*** ******

***

**

**

** **

**

*

* *

*

*

*

*

*

*

**

**

**

** **

EOS

NS

FFA

Chol

Figure 3 | Impaired acylceramide formation in Pnpla1 / epidermis (a) Representative TLC analysis of lipids extracted from Pnpla1þ / þ, Pnpla1þ /  and Pnpla1 / epidermis In Pnpla1 / mice, EOS, GlcEOS and OAFHA were almost completely depleted (blue), with reciprocal increases in OH FA,

o-OH Cer and GlcCer (red), relative to Pnpla1þ / þ and Pnpla1þ /  mice TG, triglyceride; FFA, free fatty acid; Chol, cholesterol; PE,

phosphatidylethanolamine; PC, phosphatidylcholine: SM, sphingomyelin (b–e) LC–MS/MS analysis of epidermal ceramide and related lipid species showing marked reductions in EOS (b) and OAHFA (d) species with linoleic acid (18:2) and increases in corresponding o-OH Cer (c) and o-OH ULCFA (e) species in Pnpla1 / mice in comparison with Pnpla1þ / þmice (mean±s.e.m., n¼ 3 animals; *Po0.05, **Po0.01 and ***Po0.001 versus Pnplaþ / þ

mice) Inb,c, C18-sphingosine-based ceramide (d18:1) species are selected and shown Results from one or two independent experiments are shown

(Supplementary Fig 5e), probably because of the perturbed LA

metabolism resulting from impaired formation of OLHFA and

acylceramide.

Keratinocyte-specific Pnpla1 ablation impairs skin barrier To

ascertain whether the skin barrier defects observed in global

Pnpla1 /  mice were indeed intrinsic to skin, mice carrying the

loxP-flanked Pnpla1 allele (Pnpla1f/f) were crossed with mice

transgenic for Krt14 promoter-driven Cre recombinase to obtain

mice lacking PNPLA1 selectively in epidermal keratinocytes

(Pnpla1f/fK14-Cre) Expression of Pnpla1 in the skin was reduced

by B80% in Pnpla1f/fK14-Cre mice in comparison with control

Pnpla1f/fmice (Fig 4a), confirming that Cre-mediated

recombi-nation efficiently ablated Pnpla1 in epidermal keratinocytes.

About half reduction of Pnpla1 expression was also evident in the stomach, in which the K14 promoter is active46, yet it is unlikely that this reduction could influence the skin phenotype since global heterozygous Pnpla1þ /  mice showed no abnormality Although Pnpla1f/fK14-Cre animals were indistinguishable from control littermates shortly after birth, the mutant mice died within 6 days (Fig 4b) The death was accompanied by focal desquamation with a markedly elevated TEWL value, whereas the value in other unaffected skin region remained unchanged (Fig 4c,d) Histologically, a lower zone of the SC layers became densely packed with lipid-poor interspaces in Pnpla1f/fK14-Cre mice at P5 (Fig 4e), as was seen in global Pnpla1 /  mice (Fig 1f) Immunostaining of Pnpla1f/f K14-Cre mice skin demonstrated diminished expression of filaggrin and loricrin in comparison to Pnpla1f/fmice (Fig 4f) Furthermore, epidermal

f/f K14-Cre

f/f

f/f

f/f f/f

f/f K14-Cre

f/f K14-Cre

f/f K14-Cre f/f f/f K14-Cre

f/f K14-Cre f/f f/f K14-Cre

1.5 1 0.5 0

Brain

TG Chol EOS NS

GlcEOS GlcCer

PE PC ω-OH Cer

Lung Liver Kidne y Spleen Hear t StomachIntestine

Skin

100

75 50 25 0

(day)

15

10

5

0

10 8 6 4 2 0

**

**

**

*** ***

b a

Filaggrin/DAPI

Loricrin/DAPI

—

Figure 4 | Phenotypes of keratinocyte-specific Pnpla1-deficient mice (a) qPCR analysis of Pnpla1 expression in various tissues of control (f/f) (n¼ 2) and Pnpla1f/fK14-Cre (f/f K14-Cre) (n¼ 4) mice at P5 (b) Postnatal death within 6 days due to epidermal-specific disruption of Pnpla1 (n ¼ 5 per genotype) (c) Gross appearance of control (f/f) and mutant (f/f K14-Cre) mice at P5 Mutant animals showed smaller body size Yellow circles labelled with a,b indicate regions without and with severe desquamation, respectively (d) TEWL of control (n¼ 13) and mutant (n ¼ 18) mice at P5 Labels a,b are as indicated inc (e) Representative images of hematoxylin-eosin staining of skin sections from control and mutant mice at P5 In mutant mice, the lower part

of the SC layers became densely packed with poor lipid interspaces (arrow) (f) Impaired terminal differentiation of epidermal keratinocytes at P5 in Pnpla1f/fK14-Cre mice Sections were stained with anti-filaggrin and anti-loricrin antibodies (red) and DAPI (blue) (g) Densitometric analysis of TLC separation of epidermal lipids extracted from control and mutant mice at P5 Individual lipid levels were normalized with SM content (n¼ 6 animals)

Ina,d,g, values are mean±s.e.m.; *Po0.05, **Po0.01 and ***Po0.001 versus control mice Scale bars in e,f, 20 mm Data are from at least two independent experiments

levels of EOS and GlcEOS were markedly lower, while those of

o-OH Cer and GlcCer were conversely higher, in Pnpla1f/fK14-Cre

mice (Fig 4g) Taken together, these results suggest that PNPLA1

is required in a cell-autonomous manner for acylceramide

formation and keratinocyte differentiation.

EOS rescues aberrant differentiation of mutant keratinocytes.

To further investigate the function of PNPLA1 in keratinocyte

differentiation, gene expression in primary keratinocytes

prepared from Pnpla1 /  and control mice was analysed in

culture Consistent with the in vivo data (Fig 2a–c), expression of

the terminal differentiation marker Flg was lower in differentiated

Pnpla1 /  keratinocytes than in replicate control cells, while

that of Ppard or Hbegf was significantly elevated in differentiated

Pnpla1 /  keratinocytes (Fig 5) Supplementation of the

differentiation medium with EOS(C30:0) partially reversed the

altered expression of Flg, Ppard and Hbegf in Pnpla1 / 

kera-tinocytes (Fig 5) These results suggest that the PNPLA1 product

EOS or its derivative(s) regulates terminal keratinocyte

differ-entiation partly through modulating PPARd expression.

Epidermal lipid composition in Pnpla1 /  and Abhd5 / 

mice Last, we compared epidermal lipid composition between

mouse lines deficient in Pnpla1 and Abhd5 (Cgi58, a co-factor for a

putative TG lipase), both of which appear to converge on the

processing of acylceramides22 Abhd5 /  epidermis at E18.5

showed partial reductions of OAHFA and EOS, an almost total

depletion of GlcEOS, substantial increases of o-OH Cer and

GlcCer, and marked accumulation of TG (Fig 6a–c) These lipid

profiles in Abhd5 /  mice were similar to those in Pnpla1 / 

mice, except that TG accumulation was not evident in the latter.

Although the increase in TG content in Pnpla1 /  epidermis at

P0 may be explained by the induction of lipogenic enzymes

(Fig 3a; Supplementary Fig 3a), the distinct impact of Abhd5 and

Pnpla1 ablations on TG levels at E18.5 lends further support to

segregation of PNPLA1 from bulk TG hydrolysis in which ABHD5

participates Nonetheless, the similar reductions of OAHFA, EOS

and GlcEOS in both Abhd5 / and Pnpla1 /  mice support the

cooperative roles of ABHD5 and PNPLA1 in the process of

o-O-esterification; ABHD5 in assisting the bulk release of FFAs

including LA from TG in lipid droplets and PNPLA1 in esterifying

part of this LA pool into the free and/or ceramide-bound forms of

o-OH FAs as an acyltransferase or transacylase Overall, our

results provide unequivocal evidence that PNPLA1 is a

long-sought enzyme responsible for o-O-esterification in acylceramide

biosynthesis leading to proper formation of SC lamellae,

keratinocyte differentiation, and thereby skin barrier function

(Fig 6d).

Discussion

It is generally known that PLA2 is a group of enzymes that hydrolyse the sn-2 position of glycerophospholipids to give rise to fatty acids and lysophospholipids In fact, by hydrolyzing glycerophospholipids, cytosolic PLA2a plays a central role in arachidonic acid metabolism in a wide variety of cells, secreted PLA2s modulate tissue-specific homoeostasis or diseases in given extracellular microenvironments, and PNPLA9 (iPLA2b) and PNPLA8 (iPLA2g) participate in energy metabolism and neurode-generation26,47 However, it has recently become obvious that several members of the PNPLA/iPLA2family catalyse forms of lipid metabolism other than the typical PLA2reaction, as exemplified by PNPLA2/ATGL (iPLA2z) acting as a major TG lipase in lipolysis and PNPLA3 (iPLA2e) probably acting as an acyltransferase or transacylase leading to TG accumulation in non-alcoholic fatty liver disease24,48 Herein, as part of our ongoing attempts to clarify the biological roles of the PLA2 family using comprehensive gene targeting strategies, we have identified PNPLA1, which represents

an ichthyosis-causative gene with unknown function29, as an enzyme essential for the biosynthesis of acylceramide, a unique lipid component, the presence of which has long been recognized

as prerequisite for normal skin barrier function.

Three abundant lipid groups were almost completely absent in Pnpla1 /  epidermis One of these groups is the acylceramide EOS (and EOP), a key lipid intermediate that is an absolute requirement for formation of the skin barrier and contains saturated, monounsaturated or diunsaturated ULCFA in the N-acyl chain and linoleate in the o-O-N-acyl chain The second group is GlcEOS, a glucosylated form of EOS, which can be stored in lamellar bodies to be secreted into the intercellular space of the SC and then converted back to EOS by the glucosidase GBA The third group is linoleate-containing OAHFA (OLHFA), as described below The corresponding accumulation of putative precursors of these three lipid groups, namely OH Cer, OH GlcCer and

o-OH ULCFA, in PNPLA1-deficient epidermis provides strong evidence that PNPLA1 acts as an o-O-acyltransferase or transacylase required for acylceramide synthesis In this regard, the accompaning study by Ohno et al.49has clearly shown that exogenous overexpression of PNPLA1 in cells or PNPLA1-reconstituted proteoliposomes promotes acylceramide formation likely as a transacylase and that PNPLA1 mutations associated with ARCI inactivate this transacylase activity.

So far, the order and molecular mechnisms by which ULCFAs and specifically LA are hooked onto the o-OH ULCFAs of (glucosyl)ceramides has not been fully clarified15,21 Our new proposed model for epidermal ceramide metabolism is as follows (Fig 6d): LA is directly tranferred from a linolate-containing

TG pool to the o-OH ULCFA moiety by PNPLA1 as a

20

–2)

–2)

–1)

10

0

20

10

0

Control EOS

20

10

0

Figure 5 | Cer EOS partially rescues aberrant differentiation of Pnpla1 /  keratinocytes qPCR analysis of gene expression in Pnpla1-deficient and control keratinocytes treated with 1.2 mM CaCl2for 48 h EOS was added to the culture medium at 10 mM for the last 24 h Data are presented as the mean±s.e.m (n¼ 4; **Po0.01 and ***Po0.001) Results are representative of two experiments

CoA-independent transacylase to form OAHFA, EOS and/or

GlcEOS Glucosylation of ceramide occurs in the cis-Golgi

apparatus through the action of UDP-glucose ceramide

glucosyltransferase, UGCG, and then the resulting GlcEOS is

incorporated into lamellar bodies and secreted into the

intercellular space of the SC At the SC interstices, the glucosidase GBA deglycosylates GlcEOS to EOS, which form lipid lamellae together with cholesterol and FFA Two lipoxygenases, ALOX12B and ALOXE3, sequentallly oxygenate the linoleate moiety in EOS (ref 50), which then allows a second

TG

TG

TG VLCFA

HO

HO

O O

O

O O

O

O HO

O

O

O

HO

HO

O

O

O

O

O

Keratinocyte terminal differentiation

OH

OH

OH

Lamellar membrane

ALOXE3,ALOX12B,

Unknown lipase, TGM1

CLE

LA PNPLA1

OLHFA

EOS

OH

O OH

EOS

Other FAs ABHD5/CGI-58 Unknown lipase

10

5

0 15

TG OAHFA

OAHFA

OAHF A

FFA Chol ω-OH FA

ω-OH FA

ω-OH F

A ω-OH Cer

ω-OH Cer

ω-OH Cer-CE

ω-OH ULCFA-CE ω-OH Cer-CE

ω-OH ULCFA

ω-OH GlcCer ω-OH Cer ULCFA (C≥28)

ω-OH Cer GlcEOS

Glc EOS

GlcEOS GlcCer

ω-OH GlcCer

ω-OH GlcCer

ω-OH GlcCer GlcCer

WT

c

Abhd5–/–

Abhd5 –/–

Pnpla1–/–

ELOVL4

CYP4F39

ABCA12 GBA

CERS3

Pnpla1 –/–

EOS

EOS

EOS NS

1

2

3

*** ***

***

***

***

***

***

***

***

***

**

d

Figure 6 | Roles of PNPLA1 and ABHD5 in epidermal ceramide metabolism (a) Representative TLC analysis of lipids extracted from WT, Abhd5 / , and Pnpla1 / epidermis at E18.5 (b) Quantification of panel A by densitometric analysis (mean±s.e.m., n¼ 6, 5 and 5 for WT, Abhd5 / and Pnpla1 /  mice, respectively; **Po0.01 and ***Po0.001 versus WT mice) Cumulative results of three independent experiments are shown (c) A summary profile for a,b Up and down arrows represent an increase and a decrease, respectively, in the level of individual lipids in mutant mice relative to WT mice The number of arrows indicates the relative degree of difference (d) Schematic diagram showing the proposed role of PNPLA1 in epidermal ceramide metabolism in association with keratinocyte differentiation and skin barrier function Significantly increased and decreased lipid metabolites in Pnpla1 / epidermis are highlighted in red and blue, respectively Upregulated enzymes are shown in italics LA derived from TG appears to be esterified at the o-position of o-OH ULCFA, o-OH Cer and/or o-OH GlcCer (reactions 1, 2 and 3, respectively) by PNPLA1 transacylase For details, see text

as-yet-unidentified lipase to de-esterify acylceramides The

resulting pool of o-OH Cer can then be covalently linked to

the outer suface of the CE, thus foming the CLE.

Although it has been reported that OAHFAs are abundant in

the epidermis51, the function and origin of this unique class of

lipids in the epidermis have remained unknown It is likely that

OAHFAs share a biosynthetic reaction with acylceramides, which

also contain an N-acyl chain composed of a particular type of

OAHFA, namely linoleate-containing OLHFA There are at least

two possible pathways for OAHFA biosynthesis, either directly via

o-O-esterification of o-OH ULCFA with LA by PNPLA1 (Route 1

in Fig 6d) or indirectly via synthesis of acyl(glucosyl)ceramides by

PNPLA1 and subsequent hydrolysis by a ceramidase (Route 2

and 3 in Fig 6d) Whether CERS3 could use OAHFA (or its -CoA

form) as a substrate for acylceramide synthesis remains to be

determined.

The marked alteration of epidermal structure and function

along with impaired acylceramide synthesis in Pnpla1 / 

newborns, accompanied by down-regulation of CE proteins and

up-regulation of EGF ligands, indicate that acylceramide

biosynthesis is required for not only the water-impermeable

intercellular lipid lamellae in the SC, but also the

proper transition from proliferation to terminal differentiation

of keratinocytes The delayed onset of skin phenotypes in

Pnpla1f/fK14-Cre mice in comparison with global Pnpla1 / 

mice may be due to incomplete deletion of cutaneous

Pnpla1 expression at birth in the former The increased

expression of PPARd in Pnpla1 /  epidermis could explain, at

least in part, the induction of a panel of lipid metabolism-related

genes associated with ARCI Indeed, PPARd contributes to

up-regulation of ABCA12 and GBA in keratinocytes, and

PPARd deficiency decreases lipid metabolism required for

lamellar membrane formation and thereby skin barrier

function52–54 On the other hand, hyperactivation of

PPARd enhances keratinocyte proliferation through inducing

HB-EGF (ref 43), an event that is recapitulated in Pnpla1 / 

keratinocytes.

It is tempting to speculate that the increased extracellular levels

of acylceramide or its derivative(s) at the SG/SC border

could provide a critical signal for keratinocyte maturation to

corneocytes In our study using cultured Pnpla1 / keratinocytes,

the supplementation with EOS reversed the decreased expression

of filaggrin and increased expression of HB-EGF towards normal

levels In support of this observation, application of synthetic

pseudo-acylceramide or GlcEOS recovers diminished barrier

function in vivo and promotes maturation of cultured

keratino-cytes by facilitating cornification and CE formation55,56 Moreover,

markers for keratinocyte proliferation and differentiation are

dysregulated in several other knockout mouse lines deficient in the

pathway leading to EOS synthesis, processing or transport (for

example, Elovl4 / , Cers3 / , Abhd5 / , Abca12 /  and

epidermal-specific Ugcg / )22,35,41,57,58 In contrast, keratinocyte

differentiation is not profoundly affected in Alox12b /  mice59,

where protein-bound lipids, but not free ceramides including EOS,

are decreased, consistent with the view that the LOX-catalysed

oxidation of the linoleate residue in acylceramide is required for

subsequent ester hydrolysis and covalent binding of the resultant

free o-OH Cer to the CE50 These differences could be explained if

differentiated keratinocytes have the ability to sense an

extracellular pool of acylceramide or its derivative(s) through a

putative receptor, transporter or other way Nonetheless, the

existence of such cross-talk between acylceramide metabolism

and transcriptional control of keratinocyte differentiation would be

advantageous for the coordinated formation of corneocytes

and intercellular lamellar membranes that comprise the SC

with competent permeability barrier function, although full

understanding of the underlying mechanism needs further elucidation.

Overall, our analyses of epidermal lipids, morphology and permeability barrier function lend strong support to the contention that PNPLA1 is essential for acylceramide synthesis and skin barrier function Our genetic approach using knockout mice and the biochemical approach by Ohno et al.49complement each other

by providing different lines of evidence that prove that PNPLA1 catalyses the o-O-esterification in acylceramide biosynthesis While our manuscript was under final review, Grond et al.60 also reported that acylceramide biosynthesis was impaired in the skin of another Pnpla1 /  mouse strain and in human keratinocytes with PNPLA1 mutation, and that topical application of epidermal lipids from WT mice to Pnpla1 /  skin promoted rebuilding of the CLE Herein, by means of comprehensive lipidomics, global gene profiling and conditional targeting, we have provided additional insights that the action of PNPLA1 is highly linoleate-selective and keratinocyte-intrinsic Indeed, o-O-acyl linoleate in acylceramides and OAHFAs is largely abolished with only partial replacement by other fatty acids

in the Pnpla1 /  epidermis, implying that PNPLA1 selectively utilizes linoleic acid for acylceramide biosynthesis and that the loss

of this linoleate specificity causes epidermal barrier defect Although the catalytic mechanism, subcellular localization and functional regulation of PNPLA1 still remain to be elucidated, the findings obtained from these three complementary studies altogether contribute to a better understanding of the skin barrier formation and ichthyosis development, and should be useful in providing novel therapeutic strategies for treatment of patients with skin barrier disorders.

Methods

Keratinocyte culture.Mouse primary keratinocytes were isolated as described previously61 Briefly, skins of newborn mice were treated with 5 mg ml 1Dispase (Thermo Fisher) overnight at 4 °C The epidermis was then mechanically separated from the dermis and incubated with Accutase (Nacalai tesque) for 20 min at room temperature to collect keratinocytes Human and mouse progenitors for epidermal keratinocytes were purchased from CELLnTEC and have been tested for mycoplasma by the distributor These cells were cultured in CnT-Prime medium containing 1% (v/v) antibiotic-antimycotic solution (Thermo Fisher) After reaching confluency, the cells were cultured for appropriate periods in CnT-Prime 2D Diff medium supplemented with 1.2 mM CaCl2to induce keratinocyte differentiation As required for experiments, Cer EOS (N-(30-Linoleoyloxy-triacontanoyl)-sphingosine; Matreya LLC) was dispersed by sonication for 1 min and then added to the culture All media were from CELLnTEC

Mice.Pnpla1-deficient mice, containing a ‘knockout-first’ allele targeted to the Pnpla1 genomic locus named Pnpla1tm1a(KOMP)Wtsi, were generated from a conditional targeting vector obtained from the Knockout Mouse Project resource (KOMP-CSD ID:79620) (ref 62) Briefly, mouse embryonic stem cells derived from C57BL/6N mice (RENKA)63containing the correctly targeted Pnpla1 locus were injected into blastocysts and transplanted in pseudopregnant mice to generate chimaera mice Highly (80–90%) chimeric males were mated with C57BL/6N females, and germ line transmission of the targeted allele was confirmed by PCR The IRES-LacZ and Neo cassettes were removed by flippase-mediated excision The male and female heterozygous mice were intercrossed to obtain homozygous null mice, and littermate WT mice were used as controls Mice with a floxed allele of Pnpla1 were crossed with transgenic mice for K14 promoter-driven Cre recombinase46to obtain skin-specific Pnpla1 / mice (Pnpla1f/fK14-Cre)

To generate Abhd5 / mice, genomic Abhd5 clones were isolated from mouse 129v/Ev genomic library A 6.7-kb fragment of an Abhd5 clone was subcloned into

a targeting vector with exon 1 being replaced by the PGK-Neo cassette The targeting vector was introduced into 129Sv/Ev embryonic stem cells and a correctly targeted embryonic stem cell line was injected into blastocysts, resulting in the gene-targeted mouse strain Heterozygotes were backcrossed onto C57BL/6 J background for at least five generations and then intercrossed to obtain homozygous null mice

Genotyping of offspring was performed by PCR of tail-snip DNA using genotyping primers (Supplementary Table 1) Animals were fed ad libitum (CE2, Clea Japan), had free access to water, and were kept on a 12:12-h light:dark cycle in single cages All experimental procedures involving animals in this study were approved by the Institutional Animal Care and Use Committees of Tokyo Metropolitan Institute of Medical Science and Nagoya University and were